•Dictation no longer dies the instant it starts. A guard meant to catch keyboard shortcuts was overzealous and cancelled real recordings about a quarter-second in. It's been pulled until we make it smarter.
✨ Improvements
•It hears you better now. The last release's noise filtering got a little trigger-happy and was dropping or swapping real words. We dialed it back, so what you say is what you get.
v0.9.2
June 2026
🐛 Bug Fixes
•No more made-up text from silence. Whisper used to invent sentences (sometimes in random languages) when it heard noise or nothing. Those are now detected and dropped instead of typed out.
•Cleanup stays a cleanup. The AI no longer occasionally "replies" to a transcription; if it does, we keep your original words.
•The "couldn't paste" popup dismisses for real now. It was being blocked from closing itself.
v0.9.1
June 2026
🐛 Bug Fixes
•The "couldn't paste" popup closes again. It was getting stuck open; now it always dismisses on its own (and still pauses while you hover to copy).
v0.9.0
June 2026
✨ New Features
•Whisper is built in now. Transcription runs inside the app, with the model kept warm in memory and the GPU doing the work on Apple Silicon. No more "install whisper-cpp", just pick a model.
•Much faster dictation. The model loads once instead of reloading from disk every time you talk, so everything after the first dictation is noticeably snappier.
v0.8.8
June 2026
🔧 Improvements
•Snappier and lighter. Faster decoding, less waiting before text appears, reused network connections for cloud cleanup, and a smaller download.
v0.8.7
June 2026
🐛 Bug Fixes
•Fixed a serious memory leak. Leaving Dict running and dictating used to slowly eat memory until things got ugly. The recording overlay is now reused instead of rebuilt every time, so memory stays flat.
v0.8.6
June 2026
🔧 Improvements
•Tidier menu. Dropped the "Start Dictation" tray item, dictation is the hotkey's job, not a menu click.
v0.8.5
June 2026
🐛 Bug Fixes
•Voice commands fire more reliably. A command now triggers when your phrase contains the trigger, and punctuation no longer trips it up, so "okay, take a screenshot" works.
v0.8.4
June 2026
✨ New Features
•Friendlier onboarding. A warmer opener and a "Make it yours" step to set your language and tone before you start.
•Quicker setup. Momentum-style steps so first run feels like a minute, not a form.
v0.8.3
June 2026
🔧 Improvements
•Smarter cleanup. Collapses stutters, repeats, and false starts, and drops tics like "I don't know" when they add nothing.
•Tidier history and account. The Dict Cloud badge shows on short entries with a clearer tooltip, and the usage view now shows how much you've used.
🐛 Bug Fixes
•No more stray transcriptions. If your hotkey is a single modifier, pressing another key (like Option+Arrow) now cancels the recording instead of dictating.
v0.8.2
June 2026
🔧 Improvements
•Cleaner LLM picker. Dict Cloud only shows as an option when you're signed in and it's enabled for your account.
•Live usage. Your daily usage updates on the Account page as you go.
v0.8.1
June 2026
🔧 Improvements
•Updates itself now. This is the first build that pulls future versions over the air. You install it once and forget about it.
•Windows updates too. Fixed the Windows entry in the update manifest so it stops getting left out.
v0.8.0
June 2026
✨ New Features
•Over-the-air updates. Dict checks for new versions and updates in place. No more re-downloading.
•Dict Cloud. Optional managed cleanup with accounts and plans, for when you do not want to run a model yourself. Off by default.
•Linux and Windows builds. Signed installers for all three platforms, straight from CI.
•Copy fallback. When Dict cannot paste into an app, it hands you the text in a quick popup instead of losing it.
v0.6.0
April 2026
✨ New Features
•Whisper model manager. Download and switch models from inside the app. No more hunting for files.
•Configurable hotkey. Set any shortcut you want instead of the default.
•Preferences overhaul. Cleaner sidebar, more breathing room, less clutter.
v0.5.0
March 2026
⚡ Changes
•Whisper is the only engine now. Apple Speech is gone. Same privacy, and it works the same on every platform.
•In-app editors. Edit history, dictionary, snippets, and voice commands without touching a config file.
•Privacy mode. Skip cloud LLMs and stop saving history with one toggle.
•15-second cap. Recordings stop at 15 seconds to keep things snappy.
v0.4.0
March 2026
✨ New Features
•Rebuilt on Tauri 2.0 and Rust. One codebase, ready for three platforms, lighter than the old build.
•Cross-platform groundwork. The plumbing for Linux and Windows lands here.
v0.3.3
February 2026
🐛 Bug Fixes
•Fixed dropped transcriptions: No longer preempts final result with empty delivery. Your words actually make it to the screen now.
•Error handler cleanup: Better error handling that doesn't leave things in a weird state.
🔧 Improvements
•Accessibility permission required: Now properly requests Accessibility access for reliable text injection.
•Whisper setup guide: Added documentation for setting up local Whisper.
v0.3.2
February 2026
✨ New Features
•DictCore library: Extracted core functionality into a reusable library.
•Test infrastructure: Added testing framework for better reliability.
🐛 Bug Fixes
•Push-to-talk fix: Resolved issues with push-to-talk mode.
🔧 Improvements
•Signed DMG: DMG is now properly signed for smoother installation.
v0.3.1
February 2026
🔧 Improvements
•Branded DMG: Custom installer with Gatekeeper fix script.
•Notarization support: App is now notarized by Apple for smoother first launch.
v0.3.0
February 2026
✨ New Features
•Voice Commands: Map spoken phrases to keyboard shortcuts. Say "take a screenshot" to trigger ⌘⇧3.
•Branded DMG Installer: Custom DMG with gradient background and drag-to-install arrow. Looks pretty now.
•Modern Preferences: Redesigned preferences window with sidebar navigation.
•Update Checker: Automatically checks for new versions on launch.
🔧 Improvements
•Improved text injection reliability across more apps
•Better LLM processing with refined prompts
•Overlay enhancements and smoother animations
v0.2.0
February 2026
✨ New Features
•LM Studio Support: Added LM Studio as a separate LLM provider alongside Ollama.
•Flow Mode: Auto-press Return after pasting. Perfect for chat apps and terminals.
•Overlay Position: Choose between top or bottom screen position for the recording pill.
•Code Signing: Proper code signing for persistent permissions across updates.
🔧 Improvements
•Redesigned recording overlay: dark pill with white dot-level indicators
•Processing animation stays visible during STT/LLM processing
•Lock endpoint/API key fields based on provider selection
🐛 Bug Fixes
•Fixed pill staying open on empty transcription or STT errors
v0.1.0
February 2026
🎉 Initial Release
•macOS menu-bar app that converts voice to text via global hotkey (⌥Space)
•Two STT backends: Apple Speech Recognition and Whisper (via whisper-cpp)
•Optional LLM post-processing with Ollama support
•Push-to-talk and toggle hotkey modes
•Voice snippets for text expansion
•Custom dictionary for proper noun corrections
•Privacy-first: runs 100% locally, no cloud required